diff options
author | Nadav Rotem <nadav.rotem@intel.com> | 2011-08-11 16:41:21 +0000 |
---|---|---|
committer | Nadav Rotem <nadav.rotem@intel.com> | 2011-08-11 16:41:21 +0000 |
commit | 5e742a3e1b29104fcadc26286f6daa6c450f87d9 (patch) | |
tree | ff212dc32e40b08dbbd2b39b49d0445893577ccb /test/CodeGen/X86 | |
parent | 7eba85eb86e349a3e707dd6ffa52b039fb2c28e0 (diff) | |
download | external_llvm-5e742a3e1b29104fcadc26286f6daa6c450f87d9.zip external_llvm-5e742a3e1b29104fcadc26286f6daa6c450f87d9.tar.gz external_llvm-5e742a3e1b29104fcadc26286f6daa6c450f87d9.tar.bz2 |
[AVX] If the data which is going to be saved is already in two XMM registers
(for example, after integer operation), do not pack the registers into a YMM
before saving. Its better to save as two XMM registers.
Before:
vinsertf128 $1, %xmm3, %ymm0, %ymm3
vinsertf128 $0, %xmm1, %ymm3, %ymm1
vmovaps %ymm1, 416(%rsp)
After:
vmovaps %xmm3, 416+16(%rsp)
vmovaps %xmm1, 416(%rsp)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@137308 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86')
-rw-r--r-- | test/CodeGen/X86/avx-insert.ll | 17 |
1 files changed, 17 insertions, 0 deletions
diff --git a/test/CodeGen/X86/avx-insert.ll b/test/CodeGen/X86/avx-insert.ll new file mode 100644 index 0000000..d9eae03 --- /dev/null +++ b/test/CodeGen/X86/avx-insert.ll @@ -0,0 +1,17 @@ +; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck %s + +; It is faster to make two saves, if the data is already in XMM registers. For +; example, after making an integer operation. +define void @double_save(<4 x i32>* %Ap, <4 x i32>* %Bp, <8 x i32>* %P) nounwind ssp { +entry: + ; CHECK: movaps + ; CHECK: movaps + ; CHECK: movaps + ; CHECK: movaps + %A = load <4 x i32>* %Ap + %B = load <4 x i32>* %Bp + %Z = shufflevector <4 x i32>%A, <4 x i32>%B, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> + store <8 x i32> %Z, <8 x i32>* %P, align 16 + ret void +} + |